Point of view: it’s 23:37 and you’ve just finished a nail-bitter chess match on lichess.org. The game was at times chaotic, but you feel you’ve crushed your opponent. You enthusiastically click on the Analysis board, and a series of blunders startles you out of your chair… What now?!

Learning from your mistakes is key in chess, which makes the analysis of past games a guaranteed training tool for those looking to climb that rating ladder. Frankly, the integrated game review feature of playing platforms, such as the Computer Analysis option on lichess.org, can make the process really fun and instructive.

A good place to start is using the computer’s suggestions and evaluation scores to double-check your calculations. This way, you can spot the moments in the game where you only made the best decision “out of chance”, without seeing the entire picture. Similarly, you can uncover moves you thought were good, but were actually mistakes that you got away with.

You can also spot the moments where the game turned into a bit of a roller-coaster…

Ideally, you want to never go below an equal position – that would be a comfortable win.

Another useful trick is to analyse what your opponent could have done better. This way, you can determine how aware you were of their potential counter-play. It’s good to remember that things can go your way on the board because your opponent didn’t come up with effective responses, and not because your game is error-free.

You can also skim through various alternatives early in the opening. This way, you expand your overall knowledge around that opening. You can even put yourself in your opponent’s shoes and try to come up with alternatives to their moves. Then, see what the engine thinks about your decisions and what its suggested continuations are. You might learn tricks from the engine that you can later try yourself.

However, sometimes I get a bitter taste from seeing my moves as sequences of “Mistake?” and “Blunder??” comments. Especially after a victory, when I’m expecting the know-it-all engine to agree with my moves. Let’s put things into further context and look at the diagrams below:

Stockfish rates my last move, ♜f7→♜f8 a Blunder. I played ♜f7 to get the rook out of the bishop’s attack – that’s all. Yet the computer reveals that the alternative ♜ takes f3 is a much better move.

In fact, Stockfish rates the position after ♜ taking f3 at -4.7, which means Black has a decisive advantage, whereas the move I played in the game, ♜f3, gives White a +0.9 lead. The idea behind the rook’s sacrifice on f3 is allowing the black knight to capture the pawn on d4. From d4, the knight is attacking the queen on c2. The pawn on e5 also remains undefended after the capture on d4, so the black knight on d7 could capture it. In short, black ends up in a nearly equal material situation (♘ + ♟ vs ♜), but with great dynamic play.

So I could see how my move, ♜f7, should be considered a Blunder: In one move, the evaluation dropped from -4.7 to +0.9. This translates to missing a crushing variation and opting for an inferior position instead. It looks like a move that changes the trend of the game (from + to – or the other way around) in the opponent’s favour is almost always a blunder.

---

Blunders often are one-piece giveaways, like the following situation later in the same game:

My opponent moved ♛g5, but my queen can simply take theirs, so this is a one-move queen blunder. Stockfish goes as far as to warn White that after the black queen takes the White one, the checkmate is within three moves. That’s simply because Black has an overwhelming material advantage – an extra queen and an extra rook, and those pieces create unstoppable checkmate threats, such as ♛g2#.

Unfortunately, since I was down to a few seconds on the clock, I also missed the blundered queen on g5. I played bishop to b7, which is also marked as a blunder.

Moving the bishop misses out on a one-move material gain, and indirectly on a very quick checkmate. However, I still keep a significant material advantage – a bishop for a pawn – and the score is -5.1.

So ♗b7 is a Blunder, yes, but in a different way than in the first example.

On one hand, the difference between this move and its alternative Q takes g5 is huge. Since not capturing a free piece is a huge missed opportunity, any move other than ♛ taking g5 would be a Blunder. So you would want to be quick enough to see the hanging queen instantly. It’s an extremely useful skill to have.

On the other hand, this mistake does not come from a lack of strategic understanding of the position. In fact, you could get better at spotting quick opportunities by being more focused during the game, by playing longer time control games more often, and simply by getting more experienced. There’s no reason to stress about it too much.

---

The last example reveals a Mistake from another blitz game:

My opponent’s last move, rook to f8, attempts to protect the black queen on f5 from the attack of the white one on f2. It would have been better for my opponent to avoid an exchange of queens. Since black is a pawn down, trading queens works in white’s favour as it further simplifies the game. The evaluation after ♜f8 is +4.6, so White is clearly winning.

However, here’s its alternative ♛h5:

The score still doesn’t look well for black: +2.3 – a smaller, but firm advantage for white. The e5 pawn is now a strength: at the right moment, white can start pushing it towards the promotion square.

It looks like black is facing an uncomfortable position in both variations. Playing ♛ to h5 and avoiding a trade would have given him better practical chances. Therefore, ♜f8 is a mistake, not a blunder.

---

Inaccuracies are usually moves that have the right idea but are either badly executed or badly timed. They don’t change the trend of the game. An inaccuracy makes a bad position “slightly worse”. If you were in a winning situation and made an inaccuracy, you still played a good move but missed a more efficient way to convert the position into a victory.

I hope I provided some interesting insight into how Lichess judges your moves in the post-game analysis, and how you can use the feedback to find ways to improve. So next time you find yourself frustrated with a Blunder, see if you can trace the root of your errors – that alone makes you more likely to prevent blunders in the future!

By the way, wouldn’t it be nice if the Computer Analysis feature also came with a positive feedback mechanism?! A green “Excellent move!” message could sometimes sweeten the sight of the red “Blunder??” one…

It’s an idea for Lichess to think about. In fact, it has been brought up as a feature request on their blog. 

---

---

The next section covers code sections of the algorithm that sort moves into blunders, mistakes and inaccuracies. The snippets provided are written in Scala 3 and originate in the official Lichess GitHub repository. If you’re not a fan of the JVM, just skip this section!

The three possible labels – Blunder, Mistake and Inaccuracy, are expressed through the object Judgement as below:

The instances of object Judgement are later used to construct a list of winning
chances, that sets score drops thresholds to classify the three types of “judgements”:

The use of this list is best showcased in the snippet below, which is in the main body of
the instantiating method of the object CpAdvice:

Check out the code yourself.

---

---